add driver.dev benchmark results

[Brenden2008]

Hey guys,

I'm one of the engineers at driver.dev, a browser automation platform. We achieved 89 percent on these benchmarks and wanted to add our results to the repo.

Thanks!

---

Summary by cubic

Add driver as a browser provider and include Driver.dev’s Stealth Bench V1 results, with updated plots. Also improve run_eval.py to select benchmarks and rerun specific tasks.

• New Features

  • Added browsers/driver.py with session create/delete via CDP_PROXY_URL and DRIVER_API_KEY.
  • Enabled driver in browsers/__init__.py and CLI “Available browsers” list.
  • Added --bench flag to run BU_Bench_V1 or Stealth_Bench_V1.
  • Added --task-ids to rerun specific IDs or ranges; --tasks now means “first N tasks”.
  • Results saved as stealth_bench/official_results/Stealth_Bench_V1_browser_driver_model_bu-2-0.json.
  • Updated official plots and stored previous versions as *_old.png.

• Migration

  • Add DRIVER_API_KEY= to .env (optional CDP_PROXY_URL overrides, default https://bu-compat.driver.dev).
  • Example runs:
    • uv run python run_eval.py --bench stealth --browser driver
    • uv run python run_eval.py --task-ids 2 5 12 or --task-ids 29-35

^{Written for commit 990215b. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 13 files